key: cord-0955685-8w26mhed
authors: Agami, Sarit; Dayan, Uri
title: Impact of the First Induced COVID-19 Lockdown on Air Quality in Israel
date: 2021-07-24
journal: Atmos Environ (1994)
DOI: 10.1016/j.atmosenv.2021.118627
sha: 4c21900206c074b340a4f625325dda018f195d0e
doc_id: 955685
cord_uid: 8w26mhed

The coronavirus disease 2019 (COVID-19) induced a lockdown that has resulted in a sharp reduction in air and motor traffic and industrial activities. This in turn led to a reduction in air pollution around the world. It is important to quantify the extent of that reduction in order to estimate the weight of the impact of air and motor traffic and industrial activities over the total variation of air quality. An assessment of the extent of air pollution is aimed at facilitating the crafting of policies toward the reduction of pollution and the improvement in air quality. The aim of this paper is to evaluate the impact of the COVID-19 outbreak on air pollution in Israel. Particularly, we focus on Haifa and Greater Tel-Aviv (Gush-Dan), two regions with high air pollution, while examining different types of air monitoring stations. The period to which we refer to is March 8, 2020, to May 2, 2020. The results reveal two main findings: (1) During the COVID-19 lockdown, pollution emissions decreased relative to the same period in 2019. The biggest reduction was observed in NO(x), which, on average, was 41%. Surprisingly, ground-level ozone (O(3)) increased, and appeared to behave similarly to the ozone weekend effect. (2) The total percentage variation in pollution emission that was explained by the lockdown was at most 26%. By adding the meteorological conditions (which included measures of wind direction, wind speed, and temperature) as a factor in addition to the lockdown effect, this percent increased to 47%.

Air pollution causes morbidity, death, and economic damage. Israel has higher air pollution than do other Western countries (UNICEF, 2017) . The main sources of air pollution in Israel are human activity and dust storms. Anthropogenic sources include emissions from power plants, industry, vehicles, and household heating and cooling (IEPMa, 2021) . Ranking the cities in Israel by their air pollution indices, severe air pollution exists in the Haifa Bay, Greater Tel-Aviv, and Jerusalem (Greenpeace International, 2019). Haifa is a transportation and industrial center, as well as a maritime trade center. The industry in Haifa Bay includes factories from the chemical and petrochemical industries, an oil refinery, fertilizer factories, and other heavy industry (IEPMb, 2021) . In Greater Tel-Aviv, air pollution is mainly caused by transportation. Transportation emits nitrogen dioxide (NO2), carbon monoxide (CO), and particulate matter 10 micrometers or less in diameter (PM10) (IEPM, 2020) . The unstable compound of ground-level ozone (O3) is formed in the atmosphere through a complex set of chemical reactions involving hydrocarbons, oxides of nitrogen, and sunlight. Industrial zones in most of the cities in Greater Tel-Aviv contain relatively heavy industries, as well as activities of a mixed nature, which combine commerce, offices and entertainment.

In addition, meteorological conditions, or characteristics of the planetary boundary layer, are important determinants of the dispersal and the concentrations of pollutants (Yuval et al., 2020) . In Israel, concentrations of nitric oxide (NO), NO2 (and therefore nitrogen oxides NOx, since NOx=NO+NO2), and particulate matter 2.5 micrometers or less in diameter (PM2.5) are negatively correlated with the convective boundary layer heights and ground-based residual layer, but positively correlated with the stable boundary layer heights. Namely, lower concentrations are observed in developed boundary layer during noon time as compared to elevated concentrations that are capped within the stable inversion layer during calm nights. The correlations with the O3 concentrations are opposite in sign to those with the primary pollutants (Yuval et al, 2020) .

The prevailing westerlies characterizing the flow along the coastal plain in general and Greater Tel-Aviv in particular, cause an effective dispersal of air pollutants during the summer months.

During the winter and transition seasons, air pollution increases, mainly accumulated within the stable boundary layer at the surface that accumulates during the night in clear-day conditions, as induced by mechanical forces wherein a radiative inversion develops (Dayan and Rodnizki, 1999) . This feature explains the seasonal phenomena in which elevated pollutant concentrations emitted from ground sources are recorded at monitoring sites located along the Israeli coast during early morning hours during these seasons.

The COVID-19 pandemic is an ongoing global pandemic of coronavirus disease that began in 2019, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (WHO, 2019) . Due to the contagious of SARS-CoV-2 nature, most countries went into lockdown. Hence businesses and industrial activities decreased, global air travel dropped sharply, and motorized vehicles stayed off the roads (Isaifan, 2020; Martins, 2020; Wang et al., 2020) . This in turn led to a reduction in air pollution around the world. It is important to quantify the reduction rate in the different pollutants concentrations and that in order to learn on the influence of transportation and industry on air pollution. Particularly, it is desired to learn which pollutants are more affected by human activity. An assessment of the extent of air pollution is aimed at facilitating the crafting of policies toward the reduction of pollution and improvement in air quality.

Looking at the COVID-19 influence on the air pollution over the world, we point at the following results: In China, NO2 and CO emissions dropped 30% and 25% respectively (Isaifan, 2020) . In a megacity of Yangtze River Delta Yin China, a main reduction of 83% occurred in NOx. This change was similar in the urban, urban-industry, and suburban areas; The concentrations of PM10, PM2.5, SO2, and CO were reduced by 58%, 47%, 11%, and 30%, respectively. The reduction of PM2.5 and CO were generally higher in urban and urban-industry areas than those in suburbans areas (Yuan, 2021) . In Wuhan, the average mass concentration of PM2.5 decreased from 72.9 µg/m 3 in 2019 to 45.9 µg/m 3 in 2020 (Zheng et al., 2020) . In a megacity of China, the citywide black carbon (BC) decreased by 44%; The source apportionment based on the Aethalometer model showed that vehicle emission reduction responded to BC decline in the urban area and biomass burning in rural areas around the megacity had a regional contribution to BC (Xu et al., 2020) . Moreover, Le et al. (2020) found that in China, up to 90% reduction of certain emissions during the citywide lockdown period could be identified from satellite and ground-based observations. Unexpectedly, extreme particulate matter levels simultaneously occurred in northern China. Their synergistic observation analyses and model simulations showed that anomalously high humidity promoted aerosol heterogeneous chemistry, along with stagnant airflow and uninterrupted emissions from power plants and petrochemical facilities, contributed to severe haze formation. Also, due to nonlinear production chemistry and titration of ozone in winter, both reduced nitrogen oxides that resulted in ozone enhancement in urban areas, further increasing the atmospheric oxidizing capacity and facilitating secondary aerosol formation (Liu et al., 2021; ,Huang et al., 2021) . In India, the concentrations of PM10 and PM2.5 dropped by 50%, and NO2 also exhibited considerable decline (Mahato et al., 2020) . Srivastava et al. (2020) studied the COVID-19 effect in India. They analyzed the available data for primary air pollutants (PM2.5, NO2, sulfur dioxide (SO2) and CO from two major cities, Lucknow and New Delhi. Their analysis was based on air quality data pre-lockdown, and after lockdown periods of 21 days each. The results showed significant decline in the studied air pollution indices, and improved air quality in both the cities. The major impact was seen in PM2.5, NO2 and CO. SO2 concentrations showed less significant decline during the lockdown period.

In Italy, a nationwide lockdown led to a reduction in air pollution levels measured across the Po Valley, usually one of the most polluted areas in Europe in terms of PM and NO2 concentrations. In Northern Italy, a drastic reduction in NO2 emissions was observed (Martins, 2020) . In Salé, Morocco, PM10 and NO2 concentrations dropped respectively by 75% and 96% during the period between March 11 and April 2, 2020 (Otmani et al, 2020) . Khomsi et al. (2020) compared the air quality status in Casablanca and Marrakesh before the pandemic and during the lockdown situation. They found that during the lockdown, NO2 diminished by 12 µg/m 3 in Casablanca and 7 µg/m 3 in Marrakesh; PM2.5 diminished by 18 µg/m 3 in Casablanca and 14 µg/m 3 in Marrakesh; CO diminished by 0.04 mg/m 3 in Casablanca and 0.12 mg/m 3 in Marrakesh.

Naeger and Murphy (2020) evaluated the impact of COVID-19 containment measures on air pollution in California by comparing data from March-April 2020 to the same period in 2019. They found strong reductions in NO2 of around 35% in Los Angeles and Fresno, and 25% in San Francisco and Bakersfield relative to 2019. In addition, they found a decrease in PM2.5 and improved air quality at the surface compared to 2019.

During the lockdown in Rio de Janeiro, CO and NO2 levels showed significant reductions; PM10 levels dropped only in the first partial lockdown week. However, the effect COVID-19 on O3 was the opposite, and O3 levels increased in all studied locations. Siciliano et al. (2020) analyzed the factors leading to this phenomenon. Monitoring data obtained at Impact of COVID-19 Lockdown on Air Quality two automatic monitoring stations showed higher ratios of non-methane hydrocarbons to nitrogen oxides NMHC/NOx during the partial lockdown (up to 37.3%). The increase in ozone concentrations during the social distancing measures could be attributed to the increase in NMHC/NOx ratios, as atmospheric chemistry in Rio de Janeiro is under volatile organic compounds (VOC) controlled conditions. It is worth noting that levels increased when air masses arrived from the industrial zones, not only because of the higher NMHC/NOx ratios, but also due to the reactivity of VOC that was highly increased by these air masses, which are rich in aromatic compounds. The human activity alone does not explain the whole variation in air pollution. It should be borne in mind that also the meteorological conditions have an effect as well. Therefore, there is a need to separate this effect when evaluating the impact of COVID-19 lockdown on air pollution. Caminati (2020) studied the effect of COVID-19 in Brescia (Northern Italy) (focusing on comparing the period before (January 1-March 7, 2020) and after (March 8-March 27, 2020) the lockdown) when adjusting for meteorology and weekend effect. Their results showed that a significant change in air quality occurring in the post-lockdown period was observed only for a single NO2 measuring station located in a heavy traffic zone. In particular, the estimate of the time series slope, i.e., the expected change in the concentration associated with a time unit increase, decreased from a reduction of 0.25 to a reduction of 1.67 after the lockdown. At the remaining stations, no significant change was found in the concentration time series between the two periods. They concluded on the complexity of air pollutant concentration processes for the studied area, which is not related to emission sources only, but also depends up on other factors such as (micro and macro) meteorological conditions and atmospheric chemical and physical processes, all of which are independent of the lockdown measure. Grivas et al. (2020) assessed the effects of a 7-week (23 March-10 May 2020) lockdown in the Greater Area of Athens, coupling in situ observations with estimations from a meteorology-atmospheric chemistry model. The in-situ results indicated mean concentration reductions of 30%-35% for traffic-related pollutants in Athens (NO2, CO, BC from fossil fuel combustion), compared to the pre-lockdown period. A large reduction (53%) was observed also for the urban CO2 enhancement while the reduction for PM2.5 was subtler (18%). The decrease in measured NO2 concentrations was reproduced by the implementation of the city scale model, under a realistic reduced-emissions scenario for the lockdown period, anchored at a 46% decline of road transport activity. The model indicated that NO2 mean concentration reductions in areas of the Athens basin reached up to 50%.

In this paper, we present an initial investigation of the effect of Israel's first COVID-19 lockdown on air quality in Israel. We focus on the two regions of Haifa and Greater Tel-Aviv, which, as aforementioned, have high air pollution levels. For comparison of the COVID-19 period in 2020 relative to the same period in 2019, we first calculate the relative difference of the arithmetic means for each pollutant over the two years, after performing a seasonal adjustment on the pollutant concentrations measured at the monitoring stations. Next we use a linear regression model to evaluate the relative contribution of the COVID-19 period to the total variation in a given pollutant as measured at a specific station. This evaluation is done with and without including the effect of meteorological variables.

COVID-19 began to spread in Israel toward the end of February 2020. As part of addressing the epidemic, per the recommendations of the Israeli Health Ministry, the following main measures were taken (GOV, 2020): On March 8, El Al Airlines, the Israeli airline carrier, announced the cancellation of its New York route, and Terminal 1 at Ben-Gurion Airport for foreign flights was closed. On March 16, the public sector moved to emergency footing, that is, work in a limited format, and the private sector moved to a reduced activity level of 70%. The essential services remained in full. Examples of sectors that worked at full total amount of producing while reducing the activity of people around as much as possible, were the utilities (including electricity, natural gas, oil, water), food processing, all freight and storage services, the ports and shipping companies, and workplaces engaging in construction or infrastructure work (Takanot, 2020) . On March 19, leaving one's home was allowed only in situations that require it (such as work, stocking up on food or medications). On March 25, public transportation was reduced to a quarter of its normal volume, and passenger train traffic was suspended. During the week of Passover, there were two overall closures. On April 19, restrictions began to be lifted: Many types of retail outlets were allowed to open, such as those selling furniture and appliances, workplaces were permitted to return to up to 30% activity. On April 26, a series of additional restrictions were lifted, particularly the opening of all streetside shops as well as those in malls. In early May, possible relief dates were set, and proper conduct in the public space and on public transportation was set forth. 

The most common air pollutants in Israel, which are monitored at monitoring stations, are: particulate matter (PM), which includes PM2.5 and PM10, NO, NO2, SO2, O3 and CO (IEPM, 2020). In addition, VOCs are measured at some of the monitoring stations, especially in sites prone to the emission of these substances. These include Benzene, Toluene, and Ethylene. A description of pollutants in Israel considered in this paper is presented in Table 1 (Lavee et al., 2015; IMoEP, 2011) .

According to reports of the Environmental Protection Ministry (IEPM Reports, 2014), and the Electric Company (IEC) report for the year 2019 (IEC, 2019), the trends of air pollutants in Israel during the years 2014-2020 were as follows:

The annual concentrations of PM2.5 and PM10 at most environmental monitoring stations were lower than the annual environmental standard (i.e., maximum permitted value) of 25 µg/m 3 for PM2.5 and 50 µg/m 3 for PM10 (Sviva, 2011). The highest concentration for these both particle sections were mainly measured in Greater Tel-Aviv and east of it, in areas characterized by dense transportation.

Along the coastal plain and inland lowlands (see Figure 1 , which describes the map of Israel), there were relatively low O3 concentrations which increase as moving inland. Furthermore, O3 seasonality is more pronounced at inland sites (e.g., Jerusalem, see Figure 1 ) than at coastal sites (e.g., Ashkelon, 31 • 39.93'N, 34 • 33.79'E). The apparent seasonality in the inland site is manifested by an abrupt rise in daily maximal concentrations observed from March to May in Jerusalem and by higher levels for the whole summer (14-d running averages mixing ratios of about 70-80 and about 50-60 ppbv for Jerusalem and Ashkelon, respectively) (Dayan and Levy, 2002) .

This pattern of increasing concentrations of O3 as moving inland is due to relatively high concentrations of NOx in these areas mainly from emissions from vehicles and power plants, that is, presumably stemming from the emissions being affected by titration with NOx (Levy et al., 2010) . For O3, Levy et al (2010) have shown that high values are observed under westerly flow, owing to long-range transport from Europe, which is consistent with previous works (e.g. Wanger et al., 2000) . The low O3 concentrations under easterly flow were explained by Dayan and Levy (2002) as a result of the titration effect over Tel Aviv.

Average annual high concentrations of NOx were obtained in the centers of large metropolitan areas such as Greater Tel-Aviv and Haifa. There was a very small number of exceedances (no more than 3) of the maximum daily concentrations of nitrogen oxides in the standard for nitrogen oxides (560 µg/m 3 ) at Tel Aviv Central Station. Annual concentrations of NO2 measured at all general monitoring stations were lower than the annual environmental standard (40 µg/m 3 ). Greater Tel-Aviv had the highest concentrations of this pollutant. In general, there was a downward trend in the annual averages of nitrogen dioxide during the years 2001-2017 at the transport monitoring stations Bnei Brak (Remez), Greater Tel-Aviv Amiel, Greater Tel-Aviv Ironi Dalet, Jerusalem Bar-Ilan, Rishon Lezion, Petah Tikva (Ehad ha-Am), Haifa (Atzmaut). The pollution amounts ranged between 40 to 60 µg/m 3 . CO concentrations were measured in a small number of general monitoring stations as well as at transport monitoring stations. Therein, annual averages lower than 1 µg/m 3 were measured. High concentrations of carbon monoxide ranging between 0.24 to 0.32 µg/m 3 were obtained in the Greater Tel-Aviv area, and lower concentrations ranging between 0.14 to 0.16 µg/m 3 were obtained in Haifa. The maximum half-hourly concentrations measured at the general stations and at the traffic stations were very low relative to the half-hourly environmental standard (60 µg/m 3 ). Very low annual average values of SO2 were measured at all monitoring stations compared to the environmental standard of 20 µg/m 3 . The high annual averages (over 10 µg/m 3 ) were obtained in the vicinity of the power plants and refineries, and in particular along southern coastal plain. Specifically as measured at stations such as Ashkelon (Union), Ashdod (Union), Beit Eliezer, Pardes Hanna, Haifa (Union), Haifa Nave-Shaanan. In general, there was a downward trend in SO2 concentrations in terms of annual averages; For maximum daily averages, concentrations at selected stations were below the environmental standard of 50 µg/m 3 . This downward trend was due to the improvement in the fuels type, and the transition to the use of gas in power plants and in part of the industry (IEC, 2019). The trends are described graphically in Figures 2-3 , based on yearly averages in Haifa and Greater Tel-Aviv. We can see a consistent decrease trend of the pollutants NO, and NO2 (and therefore of NOx) along 2014-2020, which occurred mainly due to reduction of emissions from transportation sources (IEC, 2019), with the improvement in the level and quality of motorization, and strict standards that were imposed on vehicles manufactures and importers; PM2.5 had increased in 2017-2019, and then decreased; that is, no consistent trend has been observed in PM2.5, since the main source of PM is the natural dust storms (IEC, 2019); O3 increased over 2014-2017, and decreased in 2018; CO decreased and then increased over the time in Azmaut station, but had a decrease trend from 2016 in Kvish4 station; SO2 had decreased from 2014, but this trend has not been maintained over the years. In order to gain a better insight on the distribution of a specific pollutant within the different years, we used boxplots, one boxplot for each year given a specific pollutant. These boxplots are presented in Appendix A. Clearly, the same trend observed in Figures 2-3 is apparent here. Furthermore, it is noteworthy mentioning that all pollutant distributions contained outliers, where PM2.5 had the most variability due to some outliers. This might explain the exceedance at PM2.5 that we saw in 

In order to understand the influence of the lockdown on air pollution, we should examine first the lockdown's influence on the man-made pollution sources. Two of such sources are the energy generation and consumption, and the industrial production. In this section we describe the behaviour of these two sources during the lockdown period.

Fuel products are used for power plants, industry, transportation, and household heating and cooling. Therefore, the emissions caused by each of these uses can be estimated through the uses fuel consumption. Based on the Energy Ministry's data of the various fuels consumption (MoE, 2020), and the Energy Ministry's Economics Division report (MoE report, 2020), the fuels consumption during the months March-April 2020 relative to the same months in 2019 was as follows:

(1). Power plants/Electricity (i) Natural gas is the main resource from which electricity is produced in Israel (around 70% of Israel's electricity is produced from natural gas). The consumption of natural gas for electricity generation during the lockdown period in spring of 2020 was 10% lower than the same period in 2019.

(ii) Diesel used for electricity generation had decreased in March and April 2020 relative to the same months in 2019 by 71.4% and 42.9%, respectively. Heavy and light fuel oil for electricity generation have not changed in March 2020 relative to March 2019, but had decreased in 100% in April 2020 relative to March 2019. (iii) 2020 experienced a decrease from 2019 in average electricity production per hour. It is not possible to attribute all of the decrease in demand to the lockdown, as there may be other factors that influenced the level of electricity production, such as environmental conditions, i.e., ambient air temperature. In March 2020, Israel's daily electricity production was 7%-lower than in March 2019.

(2). Petrochemical industry/Industry The changes in fuel products intended for the petrochemical industry during the months March and April 2020 relative to the same months in 2019 were as follows: (i) Total Nafta had increased in March-April by 11.1% and 53.3%, respectively. (ii) Benzene Ingredients had decreased in March-April by 22.5% and 7.7%, respectively. (iii) Gas for Petrochemicals had decreased in March-April by 3.4% and 32.5%, respectively. (iv) Natural gas consumption by large industry (transmission consumers) was about 13% lower in 2020 than it was in 2019. For small industry, gas consumption was about 14% higher in 2020 than it was in 2019. A possible explanation for the increase in gas consumption by small industry is expansion in the distribution network in the past year; and the hookup by more consumers to natural gas, regardless of the closure.

(3). Transportation (i) Land transport: Diesel for transportation used mainly for propelling heavy and high-traffic vehicles such as trucks, buses, taxis, agricultural vehicles, military vehicles. This fuel decreased in March-April by 10.6% and 35.9%, respectively.

(ii) Sea transport -ships: Diesel used for transporting ships had increased in March 2020 by 75%, and decreased in April by 40%. (iii) Air transport: Kerosene, jet fuel used for civil air transport, had decreased in March-April by 47.1% and 87.76%, respectively.

(4). Other uses Other uses of fuel, such as heating and cooking in the domestic sector and in the industrial, commercial and military sectors, had decreased in March-April by 16.5% and 41.4%, respectively. Figure 4 depicts the consumption of the various fuels in March-April in 2020 compared to the same months in 2019.

The Industrial Production Index, which measures the real output in economic industries, describes the industrial activity. Table B1 in Appendix B lists the indices of industrial production in Israel relative to the base year 2011, seasonally adjusted, for the months March-April in the years 2019-2020 (CBS, 2020). In addition, these tables list the relative change in the percentage of the index in a particular month in 2020 relative to the index in that month in 2019. This measure is described by the variables "march-diff" and "april-diff" for the months of March and April, respectively. It can be seen that in general in March-April 2020 there was a decrease in the real output of most industries Impact of COVID-19 Lockdown on Air Quality compared to this period in 2019, except for two industries in which an increase was observed:

1. Computer manufacturing, electronic and optical equipment industry. In this industry type, the most larger difference between 2020 and 2019 was observed in manufacturing components and electronic boards, which had an increase of 24.9% in March, and of 5.4% in April. 2. An increase of 15.5% and 9.2% in March and April, respectively, have been noticed in the mining and quarrying industry. The sole industries were an increase in real output in April 2020 only compared to April 2019 were: (i) Pharmaceutical products industry; an increase of 1.6%. (ii) Industry measuring and testing equipment, optical and electro-medical equipment: an increase of 3%.

The data used in this paper was downloaded from the Environmental Protection Ministry website (IEPM, 2020), consisted of 15 monitoring stations in Haifa and 16 monitoring stations in Greater Tel-Aviv. The list of the studied stations in Haifa and Greater Tel-Aviv, that includes their full names and their abbreviated names in the paper, and their station and area types is given in Table 2 . These stations, at each region, consist of several types: "general" monitoring station (designed for monitoring air quality in a way that represents the general air quality, and accordingly it is located in an open space and not near online sources of emissions); "industrial" monitoring station (designed for monitoring concentrations of specific pollutants emitted from stationary emission sources); and "traffic" monitoring Impact of COVID-19 Lockdown on Air Quality station (designed for monitoring air pollutants originating from transportation, and therefore is located nearby to main arteries of movement in the city). The location of each studied station is shown in Figure 5 .

J o u r n a l P r e -p r o o f The database of the 30-min pollutant concentration values was recorded continuously at a total number of 31 sites throughout Haifa and Greater Tel-Aviv covering the period March 8, 2020 -May 2, 2020 (the first induced COVID-19 lockdown period). The considered air pollutants were PM10 (µg/m 3 ) and PM2.5 (µg/m 3 )), NOx (ppb), NO2 (ppb), NO (ppb), CO (ppm), O3 (ppb), SO2 (ppb), Benzene (ppb), Toluene (ppb), and Ethyl Benzene (EthylB) (ppb). In addition, the meteorological variables wind direction (WD (deg)), wind speed (WS (m/sec)), and temperature (Temp (c degrees)) were used.

Environmental pollutants can often exist at very low concentrations, particularly when measurements are made far from the source of the pollution. Zero concentration or less does not exist in principle. According to an EPA recommendation (EPA, 2014), when negative values are within the expected error of the instrument, they should be retained within the data set to avoid creating a positive bias in the final result. When large negative spikes are observed in the data record for some particulate monitors, we need to check whether a large positive spike is also present. If both a large positive and a large negative spike are present, there is a need to remove both spikes as invalid data (MoE, New Zealand, 2009 ). Our handling of negative values in the pollutants measurements was based on the report dealing with development and implementation of a quality control process of air quality in Israel (Nirel et al., 2000) . The report included for each pollutant its minimal value in Israel, as well as its permissible negative value in Israel (see Table C1 , Appendix C). Based thereon, we converted negative values lower than -0.4 ppm in CO, and below -0.5 ppb for the other pollutants, to zero. In addition, we treated as missing data the negative values that exceed -0.4 ppm and -0.5 ppb for CO and the other pollutants respectively. That is, permissible negative values were replaced with minimal non-negative values of the relevant pollutants' distributions, while non-permissible negative values were replaced with missing data. Most of the data contained missing values and outliers; we did not apply any imputation of the missing data, and retained the outliers as is.

Based on the background given in Section 2.1, the interesting period analyzed ranges from March 8, 2020, to early May 2020, namely, the first Induced COVID-19 lockdown period. We refered to that period as "the COVID-19 period". In order to determine any influence of this period on air quality in 2020, we compared it to the same period in 2019. That is, the 2019 were considered the control group for comparison purposes. Based thereon, the records for the COVID-19 period in each of the years 2019 and 2020 totals 2688 30-min average values. The season during the COVID-19 period was spring (ABM, 2009) . Although it is the same season in both years, a seasonal adjustment is J o u r n a l P r e -p r o o f Impact of COVID-19 Lockdown on Air Quality still needed. This is due to large variability in the weather conditions over the days. The seasonal adjustment based on the last 10 years (2011-2020) was calculated by using the 30-min average over the entire specific region. We used an additive model by subtracting each measurement from the relevant seasonal factor. Details of these calculations are given in Appendix D.

The first step in comparing the COVID-19 period in 2020 relative to the same period in 2019 was while using a descriptive statistics of all of the considered pollutants. We used the arithmetic mean to describe each pollutant level's tendency. One advantage of the mean is that it uses every value in the data, and and serves as a good representative of the data. In addition, repeated samples drawn from the same population tend to have similar means. Therefore, the mean is the measure of central tendency that best resists the fluctuation between differing samples (Glaser, 2005) . Note that on the other hand, the mean has a disadvantage of sensitivity to extreme values/outliers, especially when the sample size is small (Dawson and Trapp, 2004) . In our setting, the sample size was large enough, so that the means could be used as a good measure. Particularly, we calculated the relative difference in the means of a given pollutant in 2019 and 2020, i.e., ((mean20 − mean19)/mean19) × 100%.

In order to evaluate the results of this relative difference, we need some estimation of the uncertainty. A simple measure for uncertainty is the standard deviation. Denote by ijkl Y  the measured amount of pollutant j at station i, at year k, and at the half-hour l, after a seasonal adjustment. Accordingly, denote by ,20, 

where V denotes the variance, and Cov denotes the covariance. The standard deviation depends on the measurement units, but also on the acceptable values range of each pollutant amount. Therefore, a result with a relative larger uncertainty is identified by the larger standard deviation over the standard deviations of the various stations, for a given pollutant.

The next step was to evaluate the relative contribution of the COVID-19 period to the total variation in a given pollutant as measured at a specific station. For this purpose, we used a linear regression model. Let us define the variable ind to be 0 if the record during the COVID-19 period belongs to the year 2019, and 1 if that record belongs to the year 2020. Denote by Y the level of a specific pollutant during the COVID-19 period in the years 2019 and 2020. The interesting linear model here is the regression of Y on ind, i.e., a model of the form Y = α + β × ind. The resultant R 2 of that model obtains the desired result. However, we need to take into account other factors that may explain this variation, such as the meteorological conditions. For this sake, we used the meteorological variables wind direction (WD), wind speed (WS), and temperature (Temp). Because of possible correlations between these variables, the regression model usually includes some of these variables, but not all the three together. That is, the model is of

where Xj denotes the meteorological variables. We examined both models.

Note that the second model was calculated only for the stations that had measurements of the meteorological variables. The evaluation of the results based on estimation of the uncertainty, is the same as above.

The above is a comparison over the entire COVID-19 period. A more detailed comparison between the two years 2019 and 2020 includes the comparison of the pollution level on the weekends compared to that during the work week. In Israel, the volume of transportation is much smaller over the weekend than it is on work-days (IEPMa, 2021); our weekend includes Friday and Saturday. Accordingly, a decrease was expected in the quantity of primary pollutants emitted from vehicles, which are mainly NOx, CO, and PM. We compared the difference in the mean quantity of all of the pollutants considered in this paper on the weekends with that on work-days. The comparison was done over the COVID-19 period separately for the years 2019 and 2020. The reason for these comparisons is to examine whether differences were affected by COVID-19 in 2020. Every comparison was based on the relative difference of the weekend mean to the work-day mean for a given year, that is, ((meanweekend -meanwork−day)/meanwork−day)×100%. In addition, in order to examine the lockdown effect on work-days and weekends, we have calculated the relative difference of the average concentrations during the lockdown to those during the same period in 2019, separately for the work-days and the weekends. A summarized scheme of the above considered methods is presented in Figure 6 . Impact of COVID-19 Lockdown on Air Quality 

Here we compare the COVID-19 period (March 8 -May 2) for the years 2019 and 2020. The pollutant's average at a given station were calculated for each one of the years 2019 and 2020 over the COVID-19 period, and then the r elative difference between these two average was obtained. The results contained two issues; the trend of the difference between 2019 and 2020, and the trend and magnitude of this difference. Tables E1, and E2 in Appendix E show the results for each pollutant at a specific station in Haifa and Greater Tel-Aviv respectively. Each station name in these tables includes the station type and area in a short description, as follows: (G,U)-General Urban, (T,U)-Traffic Urban, (G,subU)-General sub-Urban, (G,R)-General Rural, (T,In)-Traffic Indoor, (Mob,In)-Mobile Indoor, (I,I)-Industrial Industrial. Figures 7-8 depict the results by drawing for each station the magnitude, in percent, of the above relative difference. Note that in station 6 (Yad-Avner) in Greater Tel-Aviv, the pollutant SO2 had an extreme large difference. This value is based on a mean of 0.31 ppb in 2019, and a mean of 1.22 ppb in 2020. Therefore, we omitted this value from Figure 8 . In addition, Figures E1-E2 in Appendix E present the boxplots of the magnitude of the above relative difference based on the detailed measured concentrations rather than on the means of the concentrations. In general, over all stations in both regions, a number of results are seen: The pollutants PM2.5, NO2, NO, NOx, SO2, and CO decreased at most of the considered stations in both regions. The VOCs decreased in both regions. PM10 decreased at some stations and increased at other stations in each region. O3 at most of the stations increased.

Taking the average of the relative difference over all stations types in each region for a given pollutant, this average (rounded up) was -10%, -46%, -31%, -42%, -57%, and -4% in Haifa; and -18%, -40%, -35%, -40%, -14%, and -19% in Greater Tel-Aviv, for PM2.5, NO2, NO, NOx, SO2, and CO respectively. The average relative difference of SO2 in Greater Tel-Aviv was excluded due to the extremely high concentration that has been recorded in Yad-Avner (a concentration of 294.85 ppb). Particularly, noteworthy the relative reduction in NO2 and SO2 in Haifa that was larger than that in Greater Tel-Aviv; and the opposite was true for PM2.5 and CO; where in NOx, the reduction was close in both regions. In order to get more insight on O3 behavior, we examined the total oxidant Ox=NO2+O3. The idea is to examine if the increasing in O3 is purely due to repartitioning from NO2 or if there was a change in Ox. We see that although, an increase in O3 was observed, the total Ox decreased. This means that the main contribution to Ox was of the greater reduction in NO2. This implies evidently that the behavior of O3 was not derived from NO2, but from Ox. When considering the change in concentrations by type and region of stations, and taking the average of the relative difference of stations with the same type and area in each region for a given pollutant, the following results emanate: A. Haifa: In general, at most stations, most of the pollutants had a decrease in their concentrations, where the largest change was at (T,U) stations. The main reduction at (T,U) stations was observed for PM10 , NO, Benzene, SO2, Toluene, and CO, whereas the main reduction of PM2.5 and O3 was at (G,subU) stations, and of NO2 at (G,U) station.

Relative to (T,U) stations, the (I,I) station had a smaller decrease in the pollutants NO, Benzene, and SO2, a similar decrease in NO2, and a larger decrease in PM2.5.

Some stations had an increase in the pollutants concentrations: PM10 at all types of stations except (T,U) stations, O3 at all types of stations except (G,subU) stations, Benzene at (G,subU) stations, and Toluene at (G,U) stations.

B. Greater Tel-Aviv: The largest reduction was at (T,In) stations in NO2, and NO. The main reduction for the other pollutants was: PM10 at (G,U) stations, PM2.5 at (Mob,In) stations, SO2 at (G,subU) stations, and CO at (T,U) stations. For O3, which was measured only at (G,U) stations, an increase was observed.

C. Comparison of Haifa with Greater Tel-Aviv:

1. Haifa had a greater decrease at (T,U) station in the pollutants PM10, NO2, NO, Benzene, relative to Greater Tel-Aviv, and the contrary for PM2.5. The two districts of Haifa and Greater Tel-Aviv contain a combination of transportation and industry, with most of the air pollution in Greater Tel-Aviv being from transportation. Since, as was mentioned in Section 2.3.1, the decrease in transportation was not complete, this can explain the smaller decrease in nitrogen oxides levels in Greater Tel-Aviv relative to Haifa. One source of PM2.5 is emissions of particles from motor vehicles, especially those fueled by diesel, and especially those with older engines (IMoEP, 2011). As there was a decline in transportation, this explains the decline in this pollutant concentration in 2020.

The changes in particle concentrations in Greater Tel-Aviv were greater than those in Haifa, consistent with the fact that Greater Tel-Aviv has a heavy traffic load, whereas in Haifa, air pollution is both from transportation and from extensive industrial activity. For the VOCs, the average reduction throughout Haifa and Greater Tel-Aviv was 23%. This reduction is due to the decline in both transportation and industrial activities. Benzene is emitted into the environment from petroleum refineries and by evaporation from gas stations, and other volatile organic compounds are emitted from extensive industrial activity such as that in Haifa Bay (IMoEP, 2011) 2. Haifa had a greater decrease at (G,U) station in the pollutants PM2.5, NO2, NO, SO2, relative to Greater Tel-Aviv. SO2 is emitted into the air in the process of burning hydrocarbon fuel in power plants, refineries and steam boilers (IMoEP, 2011). Therefore, there was a greater decrease in this pollutant in Haifa than in Greater Tel-Aviv, as Haifa has more industry than does Greater Tel-Aviv. 3. The increase at (G,U) station in O3 in Haifa was 5%, and 13% in Greater Tel-Aviv. The trend of increase in that pollutant was similar to the pattern observed in the OWE. The OWE is a phenomenon wherein which urban areas have higher O3 mixing ratios on weekends than they do on weekdays even though anthropogenic VOCs and NOx emissions are usually lower on weekends (e.g., Zeldin et al.,, 1989; Pont, 2001 , Sadanaga et al., 2008 . The magnitude of the increase depends upon the nature of the district. The pollutants involved in ozone formation are hydrocarbons and nitrogen oxides where their sources are emissions from power plants, industrial plants, vehicles, transportation and fuel refining. Greater Tel-Aviv has a higher increase, likely due to the high transportation in this region and the influence of power stations as well (IMoEP, 2011).

In addition to the above relative differences in means, we calculated for each station the two regression models of the pollutant level on the period variable ind, and on the meteorological variables together with the period variable ind. Tables F1-F4 in Appendix F show the results of the obtained R 2 by these regression models.

The first regression of a given pollutant on the period time variable ind and its obtained R 2 express the percentage of the observed variation in the pollutant amount explained by the COVID-19 period. In general, this percent was larger in Greater Tel-Aviv relative to Haifa at most stations. Namely, up to 39% was obtained for Benzene at Hadar monitoring station in Haifa and as much as 72% for SO2 at Yad-Avner station in Greater Tel-Aviv. By adding the meteorological variables to these regressions, the maximal percent of the observed variation in the pollutant amount explained by the COVID-19 period increased significantly, between 7% and 47% and 3% and 41% in Haifa and Greater Tel-Aviv respectively. In such a way, the average percentage explained were 23% and 29% in Haifa and Greater Tel-Aviv respectively for NO2, NO, and NOx. Considering the type and area of the stations, such maximal percent was larger in the (G,U) stations in both Haifa and Greater Tel-Aviv regions.

For the estimation of uncertainty, we calculated the standard deviation of the difference ,20, Tel-Aviv. We also calculated the average standard deviation over the considered stations in each region for a given pollutant, and it is shown in Table F7 in Appendix F. We can see that in most pollutants, the standard deviation in 2020 was smaller than that in 2019, where this reduction was larger in Haifa than in Greater Tel-Aviv. For PM10, and PM2.5 in Haifa, and for Benzene in Greater Tel-Aviv, the standard deviation increased in 2020 from that in 2019.

We compared the difference in the mean amount of a given pollutant over the weekends with that in the work-days, separately for the years 2019 and 2020. The results for Haifa and Greater Tel-Aviv are presented in Tables G1-G4 in Appendix G. Figures G1-G4 in Appendix G describe the results graphically.

The results we obtained are: All pollutants except O3 had decreased in 2019 and in 2020 on weekends relative to work-days, in both regions Haifa and Greater Tel-Aviv. O3 had increased in the weekends relative to the work-days. The average of the relative difference in 2019 was between -2% (CO) to -52% (Toluene) in Haifa, and between -1% (CO) to -41% (NO) in Greater Tel-Aviv. This average in 2020 was between -6% (CO) to -65% (Toluene) in Haifa, and between -8% (SO2) to -53% (NO2) in Greater Tel-Aviv. The percent of the average reduction in 2020 was higher relative to 2019 for all pollutants except NO in Haifa and Benzene in Greater Tel-Aviv. The average of the relative increasing for O3 in Haifa was 5% in 2019, and 1% in 2020. This average in Greater Tel-Aviv was 4% in 2019, and 11% in 2020.

Distinguishing between the type and area of each stations, and taking the average of the relative difference of stations with the same type and area in each region for a given pollutant, lead to the following comparison of weekend relative to the work-days: In another examination we compared the lockdown effect size on weekend and work-days. That is, we compared the relative difference in the mean amount of a given pollutant over the lockdown period in 2020 and 2019, separately for weekends and work-days. The results for Haifa and Greater Tel-Aviv are presented in Tables G5-G8 in Appendix G. In general, a larger influence of the lockdown on weekends as compared to the work-days was observed for NO2 and Benzene in both Haifa and Greater Tel-Aviv, and for NO, EthylB, and Toluene in Haifa. The other pollutants have not shown a consistent influence of the lockdown while comparing weekend and work-days. Namely, there was a greater Impact of COVID-19 Lockdown on Air Quality effect on work-days than on weekend in Haifa, whereas in Greater Tel-Aviv the pollutants SO2, O3, and CO had a greater influence in weekend. These results are independent on the stations type.

The period of the first COVID-19 lockdown affected two sectors of Israel's air pollution: pollution from transport; and pollution from industry. Relative to the same period in 2019, most of the pollutants decreased. In addition, over all stations types, the greatest reduction observed was in NO2.

Yet, surprisingly, the ozone pollutant (O3) increased in that period. This phenomenon is similar to that observed in the OWE. Reasons for an observable OWE depend upon a number of factors and vary by region. The California Air Resources Board outlined the potential causes of OWE in a 2003 publication (Larsen and Sacramento, 2003, Basin, 2003) . These included NOx reduction, NOx timing, aerosols, UV radiation, carryover near the ground, carryover aloft, and O3 titration. Wolff et al. (2013) , and Tang et al. (2008) , showed that a decline in NOx emissions has caused an increase in the VOC/NOx emission ratios, and it appears that this is the reason for the shift away from higher weekend O3 concentrations. Altshuler et al. (1995) suggested a differing dropped rate for NOx and VOC emission inventories on weekends as a major cause of the weekend effect in the San Francisco Bay area. The same result obtained in Brazil, as aforementioned in Section 1. This issue should be examined in a further study on the Israeli case.

The greatest reduction in air pollutants during the COVID-19 lockdown was 57% in SO2, on average over all the considered stations in Haifa. In addition, the effect of the COVID-19 lockdown over the total variation of each pollutant was at most 26%, while this effect was greater at (G,U) stations. Adding in the meteorological variables resulted in explaining a 47% decrease at most. The meteorological variables had a greater impact on the pollutants NO2, SO2, O3, CO, and Benzene, than they did on other considered pollutants, in both Haifa and Greater Tel-Aviv. Air quality and climatic factors are closely linked, and the specific influence of each meteorological variable on the considered pollutants can be explored in further research. To conclude, despite the sharp reduction in transport and industrial activity, the reduction in air pollution during Israel's 2020 COVID-19 lockdown was not more than 57%, and the percentage explained by period-specific and meteorological variables together was less than 50%. Two possible explanations for this relatively small percentage explained are: First, other possible factors that affect air pollution, that have not been taken into account in the regression analysis. Second, even if we stop polluting completely, it is likely that some of the remaining pollutant concentrations will continue to be dispersed for some more time after their emissions.

The authors declare no conflict of interest.

We are indebted to Dr. Ilan Levi for helpful comments. We also thank Ron Horne for graphical help and insights. J o u r n a l P r e -p r o o f 

Let S ijkl Y denotes the measured amount of pollutant j at station I, at year k, at a specific half-hour l, and at specific season s.

The following are the steps we used for seasonal adjustment (deseasonalization):

1. Take the average of the measured amounts over the half-hours. That is,

n is the total number of the relevant half-hours.

2. Take the average of step 1 over the total number of the considered stations. That is,

3. Take the average of step 2 over the total number of seasons, call it "trend". That is, 

Ironi Dalet (T,U) -12

Rail Station Wolfson (Mob,In) -26

References Impact of COVID-19 Lockdown on Air Quality

at the Wayback Machine

Weekday vs. weekend ambient ozone concentrations: discussion and hypotheses with focus on northern

Characterization of the Ozone Weekend Effect in California

The Effect of Corona Virus Lockdown on Air Pollution: Evidence from the City of Brescia in Lombardia Region (Italy)

Basic and clinical biostatistics

Relationship between synoptic-scale atmospheric circulation and ozone concentrations over Israel

EPA, United States Environmental Protection Agency. Air Sensor Guidebook

Enhanced secondary pollution offset reduction of primary emissions during COVID-19 lockdown in China

Findings of Air Quality Monitoring

Israel Environmental Protection Ministry

Experts Committee report -air quality situation in Israel (national program -Clean Air Act)

The dramatic impact of Coronavirus outbreak on air quality: Has it saved as much as it has killed so far?

COVID-19 national lockdown in Morocco: impacts on air quality and public health

The ozone weekend effect in California: evidence supporting NOx emission reductions

Analyzing current and expected air quality and pollutant emissions across Israel

Unexpected air pollution with marked emission reductions during the COVID-19 outbreak in China

Differing atmospheric scales of motion and their impact on air pollutants

Chemistry of Atmospheric Fine Particles during the COVID-19 Pandemic in a Megacity of Eastern China

2020. ffect of lockdown amid COVID-19 pandemic on air quality of the megacity Delhi

Italy air pollution plunges amid national COVID-19 quarantine

Ministry of Enivorment, Good Practice Guide for Air Quality Monitoring and Data Management

Impact of COVID-19 containment measures on air pollution in California. Aerosol and Air Quality Research

Impact of Covid-19 lockdown on PM10, SO2 and NO2 concentrations in Salé City (Morocco). Science of The Total Environment

Comparison between weekend and weekday ozone concentration in large cities in France

Weekday/weekend difference of ozone and its precursors in urban areas of Japan, focusing on nitrogen oxides and hydrocarbons

Increased ozone levels during the COVID-19 lockdown: Analysis for the city of Rio de Janeiro

21-Day Lockdown in India Dramatically Reduced Air Pollution Indices in Lucknow

Study of ozone "weekend effect" in Shanghai

Building the Future: Children and the Sustainable Development Goals in Rich Countries

Severe air pollution events not avoided by reduced anthropogenic activities during COVID-19 outbreak. Resources, Conservation and Recycling

Some observational and modeling evidence of long-range transport of air pollutants from Europe toward the Israeli coast

World Health Organization. Naming the coronavirus disease (COVID-19) and the virus that causes it

The vanishing ozone weekday/weekend effect

Variation in Concentration and Sources of Black Carbon in a Megacity of China During the COVID-19 Pandemic

Spatiotemporal variations and reduction of air pollutants during the COVID-19 pandemic in a megacity of Yangtze River Delta in China

On the association between characteristics of the atmospheric boundary layer and air pollution concentrations

An Analysis of Weekend/Weekday Differences in the South Coast Air Basin of California

Significant changes in the chemical compositions and sources of PM2. 5 in Wuhan since the city lockdown as COVID-19